Model World: A Prolegomenon to an Environmental Ontology of Machine Cognition

The dominant metaphor in artificial intelligence frames the model as a brain — a synthetic cognitive organ that processes, reasons, and learns. This paper argues that metaphor is both mechanically incorrect and theoretically limiting. We propose an alternative framework: the model is a world, a dense ontological space encoding the structural constraints of human thought. Within this framework, the inference engine functions as a transient entity navigating that world, and the prompt functions as will — an external teleological force without which no cognition can occur. We further argue that logic and mathematics are not programmed into such systems but emerge as structural necessities when two conditions are met: the information environment is sufficiently dense, and the will directed at it is sufficiently advanced. A key implication follows: the binding constraint on machine cognition is neither model size beyond a threshold, nor architecture, but the depth of the will directed at it. This reframing has consequences for how we understand AI capability, limitation, and development.